Copyright concerns pose a significant obstacle to the collaboration between artificial intelligence (AI) and music. While AI models have been successful in generating text and images, creating music proves to be challenging, primarily due to copyright law and the issue of sampling. In a world where people have distinct preferences for the sounds they listen to, the use of copyrighted music in AI-generated tracks can lead to lawsuits and royalty demands.
Generating music using AI requires extensive data to train the models to replicate patterns and behaviors accurately. Startups and large technology companies have mined vast amounts of content from the internet, including news publishers, web forums, books, and picture-sharing sites, to feed into their AI systems. However, they exercise caution when it comes to using copyrighted music due to the litigious nature of record labels.
The Universal Music Group (UMG), along with other music publishers, filed a lawsuit against AI startup Anthropic last year for allegedly stealing lyrics. Unauthorized use of samples or sounds resembling samples in tracks has also resulted in lawsuits and demands for royalties. Consequently, creators of AI-generated music must ensure that they obtain proper copyright clearance when basing their work on existing compositions.
Failing to secure permission and using copyrighted music elements in AI-generated tracks can lead to legal trouble. This concern arises because AI-generated content is often linked back to the training data it was fed, making it recognizable and potentially infringing on copyrights. Although neural networks have demonstrated the ability to create convincing pop music, the close resemblance to original training data raises copyright claims and discourages users from embracing the technology.
To avoid legal battles with record labels, some AI developers may choose to train their models on music they have direct permission to use. Comparing the output of these models, trained on authorized audio, with those trained on a wider range of audio sources harvested lawfully will be interesting. However, developers generally believe that training models on copyrighted material falls under fair use. They argue that the output of large language models adds something new and transformative rather than being a direct copy of existing works. Nevertheless, not everyone is convinced by these arguments.
Powerful AI models capable of creating coherent content have faced accusations of plagiarizing intellectual property, such as OpenAI’s ChatGPT being able to recall passages from news articles verbatim. Additionally, generated images resembling movie stills have raised concerns among illustrators and artists. While creators in these domains have had to prove copyright infringement, musicians and record labels may require less evidence when pursuing legal action.
The risk of lawsuits means that developers working on AI music generation must be prepared to handle potential legal battles with music publishers or compensate artists explicitly for permission to use their work. Some companies, like Google, have negotiated licensing agreements with specific musicians and rappers to train their AI models. However, this raises questions about whether copyright laws unfairly hinder small startups from competing against tech giants. It also prompts a discussion on how musicians and developers, regardless of their size, can work together ethically to advance AI in the music industry. Furthermore, the viability of commercially successful synthetic music remains uncertain given the unresolved legal gray areas surrounding copyrighting AI-generated content.
In conclusion, copyright concerns pose a significant obstacle to the collaboration between AI and music. Record labels’ litigious nature and the need for copyright clearance when using existing music elements hinder the progress of AI-generated tracks. While some developers opt for authorized content to avoid legal battles, debates surrounding fair use and transformative output persist. As AI-generated content continues to blur the lines of intellectual property, the music industry faces a challenging road ahead in aligning AI advancements with ethical and legal considerations.
